SONA - Personal Meditation

Sona creates a personal guided meditation for what you need right now and speaks it back in the voice you know best: your own.

Comment

SONA - A Personal Meditation in Your Own Voice

When you need to reset, SONA creates the practice you need right now, in your own voice.

SONA is a mobile-first web app for creating short, personal guided meditations. A user records or uploads a voice sample once, gives explicit consent, and SONA creates an ElevenLabs Instant Voice Clone tied to their account. Future practices reuse that saved voice ID, so a new meditation can be generated without asking for another voice sample.

The app adapts each practice to the selected intention, custom context, language of the voice sample, and available time. It blends the narration with ambient music and saves completed practices in the user's private library.

SONA is a wellness and reflection tool, not a medical service or a substitute for professional care.

What it does

  1. Choose an intention: Ease anxiety, Sleep better, Find focus, Restore energy, or write a custom request.
  2. Upload a voice sample or record one on the device.
  3. Validate that the recording is at least 15 seconds long and receive clear feedback for microphone, file-size, format, provider, plan, and credit errors.
  4. Choose a 5, 7, 10, or 12 minute practice and explicitly consent to voice cloning.
  5. Create and securely save an ElevenLabs Instant Voice Clone for the signed-in user.
  6. Generate a bilingual practice in the source language of the voice sample using eleven_multilingual_v2.
  7. Mix the narration with a local ambient track, then play and save the result in the private practice library.

Product experience

  • Mobile-first design with a focused home screen and custom audio player.
  • Voice is cloned once and reused for later practices.
  • Four distinct practice structures for anxiety, sleep, focus, and energy.
  • Custom intentions are routed into the closest structure while preserving the user's own words.
  • English and Russian are supported.
  • Practices are private per user through Supabase Auth, Row Level Security, and private storage.

How we used Codex and GPT-5.6

SONA was built and meaningfully extended through an iterative Codex workflow using GPT-5.6. The collaboration covered product decisions, implementation, debugging, audio quality, testing, deployment, and documentation.

Codex helped turn the product brief into a complete Next.js application, implement the consent and voice-cloning flow, create the reusable voice-profile model and Supabase integration, build responsive interactions, diagnose audio issues, add production safeguards, and verify the GitHub and Vercel deployment.

Key human product decisions included treating voice cloning as an explicit reusable consented asset, preserving the speaker's language, keeping scripts non-clinical, keeping raw samples out of SONA storage, and prioritizing a calm integrated audio mix.

Architecture

  • Frontend: Next.js 16, React 19, TypeScript, Tailwind CSS
  • Authentication and persistence: Supabase Auth, Postgres, Storage, Row Level Security
  • Voice cloning and TTS: ElevenLabs Instant Voice Cloning and eleven_multilingual_v2
  • Audio preparation and mixing: Browser Web Audio API and MP3 encoding
  • Deployment: Vercel

Privacy and consent

The user must explicitly consent before SONA sends a sample to ElevenLabs. The raw voice sample is not stored by SONA. Only the authenticated user's ElevenLabs voice ID and completed practice files are persisted. Users can replace or delete a saved voice in the app.

Built With

Share this project:

Updates